قالب وردپرس درنا توس
Home / Health / AI that coughs carefully is expected to play a major role as an early warning system for COVID-19-TechCrunch

AI that coughs carefully is expected to play a major role as an early warning system for COVID-19-TechCrunch



The asymptomatic transmission of COVID-19 is an important cause of the pandemic, but, of course, if there are no symptoms, who can tell them whether they should be isolated or tested? The MIT study found that a pattern hidden in the cough cleverly but reliably marks a person as possibly in the early stages of infection. It can provide a much-needed early warning system for the virus.

As doctors have known for many years, the sound of a cough is very enlightening. AI models have been established to detect diseases such as pneumonia, asthma and even neuromuscular diseases, all of which will change the way people cough in different ways.

Before the pandemic, researcher Brian Subirana had shown that coughing can even help predict Alzheimer̵

7;s disease-reflecting the results of an IBM study published a week ago. Recently, Subirana believes that if AI can speak so much from so few angles, perhaps COVID-19 may also be ignored. In fact, he is not the first person to think so.

He and his team built a place where people can cough, and finally assembled “the largest cough research data set we know of.” Thousands of samples were used to train AI models and recorded in open access IEEE journals.

The model seems to have detected subtle patterns of sound intensity, mood, lung and respiratory function, and muscle degeneration, so that asymptomatic carriers of COVID-19 and symptomatic carriers can identify 100% cough and 98.5% symptom. The specificity is 83% and 94% respectively, which means that it does not have a large number of false positives or negatives.

Subirana said: “We think this shows that even if you have no symptoms, when you get COVID, the way you make your voice will change.” However, he warned that although the system is good at detecting unhealthy coughs, it should not be treated. Used as a diagnostic tool for people with symptoms but not sure of the underlying cause.

I ask Subirana to be more clear on this point.

He wrote in an email: “The tool is testing features to distinguish between topics with COVID and those without COVID.” “Previous research has shown that you can also choose other conditions. You can design one that can distinguish many System of conditions, but our focus is to select COVID from the remaining conditions.”

For those who pay attention to statistics, the success rate is incredibly high and may cause some dangers. Machine learning models are excellent in many ways, but 100% is not a lot of numbers. When you start to think about other methods, it may happen by accident. There is no doubt that these findings will need to be verified on other data sets and verified by other researchers, but it is also possible that there is only reliable evidence in the cough caused by COVID, indicating that the computer listening system can easily hear the sound.

The team is working with multiple hospitals to build a more diverse data set, but if FDA approval is obtained, the team will also work with a private company to jointly develop an application to distribute the tool to update Used widely.


Source link